Video quality intelligent verification method and system for grouting construction

Abstract

The invention relates to the technical field of construction video quality verification, and relates to a video quality intelligent verification method and system for grouting construction. According to the method, based on component edge feature information in a to-be-detected video frame sequence, a component plane normal vector and a shooting angle in the optical axis direction of a camera are calculated, whether the shooting angle meets the grouting identification requirement or not is judged, and when the grouting identification requirement is met, the boundary contour of each target hole site area in each clear video frame is identified; calculating the local visibility integrity of each target hole site area, judging the overall visibility of the collected video, extracting the overall translation motion amount and the rotation motion amount between adjacent frames to form a collected video stability index, and evaluating the comprehensive quality of the collected video in combination with the overall visibility of the collected video and the collected video stability index. And outputting a video admission result based on the comprehensive quality of the collected video, ensuring that the collected video meets the technical requirements of AI intelligent identification, and improving the availability of the collected video.

Description

Technical Field

[0001] This invention relates to the field of construction video quality verification technology, and specifically to an intelligent video quality verification method and system for grouting construction. Background Technology

[0002] To ensure the traceability and quality control of the sleeve grouting construction process, video monitoring is commonly used on construction sites to record the entire grouting operation. Intelligent analysis of the video data enables the identification and judgment of grouting quality. However, the quality of the video directly determines the accuracy and reliability of subsequent AI recognition. If the video footage has issues such as incorrect shooting angles, blurry images, obstruction of key areas, or severe shaking, the recognition algorithm will be unable to effectively extract target features, leading to misjudgments or missed detections, posing a significant safety hazard to the project.

[0003] Existing technologies typically employ manual handheld shooting when capturing grouting construction videos. However, manual handheld shooting can easily result in issues such as excessively small angles, side shots, or tilted shots. This makes it difficult for subsequent intelligent recognition to extract feature information from target areas such as grouting holes and grout outlets. Consequently, the quality judgment standards become disconnected from the requirements of intelligent recognition, leading to recognition failures or an increased rate of misjudgment of grouting quality.

[0004] Meanwhile, during construction, personnel movement and equipment movement often block the grouting holes, making the grout discharge status invisible. As a result, video data during the obstruction period is still used for quality identification, reducing the reliability of the quality identification results and failing to effectively analyze the standardization of the grouting construction process. This provides incorrect video evidence for construction quality supervision, which in turn leads to subsequent disputes over grouting quality responsibility. Summary of the Invention

[0005] The purpose of this invention is to provide a video quality intelligent verification method and system for grouting construction, so as to solve the above-mentioned problems existing in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: On the one hand, the present invention provides a video quality intelligent verification method for grouting construction, including: extracting the video frame sequence to be detected from the video collected at the grouting construction site, calculating the shooting angle between the component plane normal vector and the camera optical axis direction based on the component edge feature information in the sequence, and determining whether the shooting angle meets the grouting recognition requirements.

[0007] When the grouting identification requirements are met, the sharpness and brightness uniformity of all video frame images in the video frame sequence to be detected are analyzed, and all clear video frames that meet the requirements are selected.

[0008] Identify the boundary contours of each target aperture region in each clear video frame, construct a region visibility distribution matrix through grid division, calculate the local visibility integrity of each target aperture region, and determine the overall visibility of the acquired video.

[0009] Obtain the pixel motion vector field between all adjacent frames, and extract the overall translational and rotational motion between each adjacent frame to form the stability index of the acquired video.

[0010] Based on the overall visibility and stability indicators of the acquired video, the comprehensive quality of the acquired video is evaluated, and the video admission result is output based on the comprehensive quality of the acquired video.

[0011] On the other hand, the present invention provides a video quality intelligent verification system for grouting construction, including a shooting angle acquisition module, a clear video frame screening module, a visibility determination module, a stability index composition module, and a video access evaluation module.

[0012] The modules are connected as follows: the shooting angle acquisition module is connected to the clear video frame screening module; the visibility determination module is connected to both the clear video frame screening module and the stability index composition module; and the video admission assessment module is connected to the stability index composition module.

[0013] The shooting angle acquisition module extracts the video frame sequence to be detected from the video collected at the grouting construction site. Based on the component edge feature information in the sequence, it calculates the shooting angle between the component plane normal vector and the camera optical axis direction, and determines whether the shooting angle meets the grouting recognition requirements.

[0014] The clear video frame filtering module, when meeting the grouting identification requirements, performs a clarity and brightness uniformity analysis on all video frame images in the video frame sequence to be detected, and filters out all clear video frames that meet the requirements.

[0015] The visibility determination module identifies the boundary contours of each target hole location region in each clear video frame, constructs a regional visibility distribution matrix through grid division, calculates the local visibility integrity of each target hole location region, and determines the overall visibility of the acquired video.

[0016] The stability index module acquires the pixel motion vector field between all adjacent frames, extracts the overall translational and rotational motion between adjacent frames, and constitutes the stability index of the acquired video.

[0017] The video access assessment module evaluates the overall quality of the acquired video based on the overall visibility and stability indicators of the acquired video, and outputs the video access result based on the overall quality of the acquired video.

[0018] Compared with the prior art, the present invention has the following beneficial effects: (1) Based on the component edge feature information in the video frame sequence to be detected, the present invention calculates the shooting angle between the component plane normal vector and the camera optical axis direction, determines whether the shooting angle meets the grouting recognition requirements, avoids the problem of abnormal feature extraction of the target hole area due to shooting angle deviation, ensures that the acquired video meets the technical requirements of intelligent recognition from a geometric perspective, and improves the accuracy and reliability of subsequent AI recognition.

[0019] (2) This invention identifies the boundary contours of each target hole location area in each clear video frame, constructs a visible distribution matrix of the area through grid division, calculates the local visibility integrity of each target hole location area, determines the overall visibility of the acquired video, realizes a refined quantitative assessment of the visibility of the target hole location, effectively filters out videos with hole location visibility integrity that meet the requirements, ensures the reliability of the data basis for judging the grouting construction quality, and realizes the effective identification and supervision of the grouting construction process.

[0020] (3) This invention obtains the pixel motion vector field between all adjacent frames, extracts the overall translational motion and rotational motion between each adjacent frame, and constitutes the stability index of the acquired video, thereby achieving accurate evaluation of the stability of the grouting construction video, effectively eliminating videos with screen shaking that exceeds the recognition tolerance range, avoiding the problem of disordered hole position feature extraction during AI recognition, and improving the accuracy of AI recognition of grouting construction.

[0021] (4) Based on the overall visibility and stability index of the acquired video, this invention evaluates the overall quality of the acquired video and outputs the video admission results to ensure that the acquired video meets the technical requirements of AI intelligent recognition, improves the usability of the acquired video, and provides reliable data support for the automated detection and traceability of grouting construction quality. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0024] Figure 2 This is a schematic diagram of the process for determining whether the shooting angle meets the requirements for grouting identification in this invention.

[0025] Figure 3 This is a schematic diagram of the system module connections in this invention. Detailed Implementation

[0026] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention. Furthermore, it should be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale.

[0027] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use. Techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and apparatus should be considered part of the specification.

[0028] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0029] Please see Figure 1 As shown in the example, the present invention provides a video quality intelligent verification method for grouting construction, including: S1, extracting the video frame sequence to be detected from the video collected at the grouting construction site, calculating the shooting angle between the plane normal vector of the component and the optical axis direction of the camera based on the component edge feature information in the sequence, and determining whether the shooting angle meets the grouting recognition requirements.

[0030] S2. When the grouting identification requirements are met, perform a sharpness and brightness uniformity analysis on all video frame images in the video frame sequence to be detected, and select all clear video frames that meet the requirements.

[0031] S3. Identify the boundary contours of each target hole location region in each clear video frame, construct a region visibility distribution matrix through grid division, calculate the local visibility integrity of each target hole location region, and determine the overall visibility of the acquired video.

[0032] S4. Obtain the pixel motion vector field between all adjacent frames, extract the overall translational motion and rotational motion between each adjacent frame, and form the stability index of the acquired video.

[0033] S5. Based on the overall visibility and stability indicators of the acquired video, evaluate the comprehensive quality of the acquired video, and output the video admission result based on the comprehensive quality of the acquired video.

[0034] Considering that the accuracy of the shooting angle detection directly determines the geometric reliability of subsequent AI recognition, the existing technology relies on manual hand-held shooting, which is prone to situations such as excessively small angles, side shots, and tilts, causing severe deformation of key areas such as grouting holes and grout outlets, resulting in recognition failure. It is necessary to combine the edge feature information of the component for geometric analysis, calculate the angle between the component plane normal vector and the camera optical axis, realize the quantitative determination of the shooting angle, and ensure that the acquired video meets the technical requirements of intelligent recognition in terms of geometric perspective.

[0035] Based on this, such as Figure 2 As shown, the specific implementation of determining whether the shooting angle meets the grouting identification requirements in this invention includes: S11, extracting the video frame sequence to be detected from the video collected at the grouting construction site, extracting frame images from the video frame sequence to be detected at a preset time interval (e.g., 1 second / frame, which can be adjusted by the implementer according to the video duration) to obtain all video frames to be analyzed, ensuring coverage of the key shooting stages of the entire video.

[0036] S12. Perform edge pixel detection on each video frame to be analyzed, extract the component edge line segment set in all video frames to be analyzed, perform straight line filtering on the component edge line segment set, and determine the horizontal edge direction vector and vertical edge direction vector of the component.

[0037] S13. Obtain the normal vector of the component plane through vector cross product operation, and calculate the shooting angle between the normal vector of the component plane and the camera optical axis direction by combining the preset camera optical axis direction vector.

[0038] It should be noted that the vector cross product operation is as follows: .

[0039] In the formula, Let be the normal vector of the plane of the component. Let be the horizontal edge direction vector of the component. is the vertical edge direction vector of the component.

[0040] The preset camera optical axis direction vector is usually set to The shooting angle is: .

[0041] In the formula, Let be the shooting angle between the plane normal vector of the component and the optical axis of the camera. Let the magnitude of the plane normal vector of the component be . Let be the magnitude of the camera's optical axis direction vector.

[0042] S14. If the shooting angle meets the preset grouting AI recognition shooting angle qualification judgment standard, such as θ≥30°, which meets the geometric requirements for clear extraction of grouting hole features, then the shooting angle is judged to meet the grouting recognition requirements. Otherwise, the shooting angle is judged not to meet the grouting recognition requirements, and the unqualified access result of the acquired video is output.

[0043] Preferably, in a specific embodiment of the present invention, the horizontal edge direction vector and the vertical edge direction vector of the component are determined as follows: S121, perform straight line identification on the component edge line segment set, and extract the corresponding horizontal edge line segment and vertical edge line segment of the component.

[0044] As an example, this invention can employ the Canny edge detection algorithm to detect edge pixels in the video frame to be analyzed, and extract straight line segments through Hough transform. The edge line segments are then classified into horizontal and vertical categories based on their direction angles; for example, the horizontal line segment direction angle ranges from -15° to +15°, and the vertical line segment direction angle ranges from 75° to 105°. Both the Canny edge detection algorithm and the Hough transform are existing techniques well-known to those skilled in the art and will not be elaborated upon further.

[0045] S122. Obtain the length and direction of the corresponding horizontal and vertical edge segments of the component, and filter out interfering segments based on the set dimensions of the grouting sleeve corresponding to the grouting construction. For example, horizontal edge segments whose length exceeds the set width range of the grouting sleeve and vertical edge segments whose length exceeds the set length range of the grouting sleeve are both recorded as interfering segments and discarded, retaining edge segments that conform to the set dimensions of the grouting sleeve.

[0046] S123. Perform directional consistency analysis on the filtered horizontal and vertical edge segments, and fit to generate the horizontal and vertical baselines of the component, thereby obtaining the horizontal edge direction vector corresponding to the horizontal edge baseline and the vertical edge direction vector corresponding to the vertical baseline.

[0047] It should be noted that the above-mentioned directional consistency analysis steps are as follows: First, the filtered horizontal and vertical edge segments are clustered by direction angle, and the K-means clustering algorithm is used to classify segments with similar direction angles into the same direction category.

[0048] Secondly, for each directional category, the coordinate distribution of the center point of the line segment is calculated, and the line segments of the same directional category are linearly fitted by the least squares method to generate the main axis of that directional category.

[0049] Finally, the main axis of the horizontal cluster fitting is projected onto the horizontal reference direction of the image coordinate system, and its average direction angle is calculated as the horizontal edge reference direction. The horizontal reference line of the component is generated by fitting along the horizontal edge reference direction, and the vertical reference line of the component is generated by fitting in the same way.

[0050] This invention calculates the shooting angle between the component plane normal vector and the camera optical axis direction based on the component edge feature information in the video frame sequence to be detected, and determines whether the shooting angle meets the requirements of grouting recognition. This avoids the problem of abnormal feature extraction of the target hole area due to shooting angle deviation, and ensures that the acquired video meets the technical requirements of intelligent recognition from a geometric perspective, thereby improving the accuracy and reliability of subsequent AI recognition.

[0051] Considering that video clarity and brightness uniformity directly affect the quality of subsequent hole feature extraction, existing technologies do not pre-screen video quality, resulting in blurred frames and abnormally exposed frames entering the recognition process, causing feature extraction failure or misjudgment. It is necessary to combine grayscale histogram statistics and Laplacian response analysis to achieve dual verification of clarity and brightness uniformity, and effectively filter clear video frames that meet the requirements of AI recognition.

[0052] Based on this, the specific implementation of the present invention includes: when the grouting identification requirements are met, performing a sharpness and brightness uniformity analysis on all video frame images in the video frame sequence to be detected, and selecting all clear video frames that meet the requirements.

[0053] In a specific embodiment of the present invention, the step of screening all clear video frames that meet the requirements is as follows: S21, perform grayscale processing on each video frame image in the video frame sequence to be detected, statistically analyze the brightness distribution characteristics of each video frame image based on the grayscale histogram, and determine whether there is abnormal exposure in the video frame image based on the brightness distribution characteristics.

[0054] Specifically, in one embodiment of the present invention, the mean and standard deviation of the grayscale histogram are extracted from the brightness distribution characteristics of each video frame image. If the mean of the grayscale histogram is greater than that of high grayscale pixels (e.g., 200) and the proportion of high grayscale pixels exceeds a set ratio (e.g., 70%), then the video frame is determined to have an overexposure anomaly. If the mean of the grayscale histogram is less than that of low grayscale pixels (e.g., 50) and the proportion of low grayscale pixels exceeds a set ratio (e.g., 70%), then the video frame is determined to have an underexposure anomaly.

[0055] S22. If a video frame does not have abnormal exposure, then the video frame is recorded as a video frame with uniform brightness.

[0056] S23. Perform convolution operation on all uniform brightness video frames to obtain the Laplacian response value of the uniform brightness video frames, and calculate the sharpness of each uniform brightness video frame through variance calculation.

[0057] Specifically, all uniformly bright video frames are converted to grayscale images. A Laplacian convolution operation is then performed on all pixels in the grayscale image to obtain the Laplacian response value of each pixel, forming the Laplacian response matrix of the entire grayscale image. The response values ​​of all pixels within the Laplacian response matrix are statistically analyzed to calculate the Laplacian response variance, which is used as the sharpness of the uniformly bright video frames. A larger Laplacian response variance indicates richer edge details and higher image sharpness in the video frame, while a smaller variance indicates blurred edges and lower image sharpness.

[0058] S24. Remove video frames with uniform brightness that do not meet the clarity standard, and count all clear video frames that meet the standard.

[0059] It should be noted that the Laplacian convolution operation is a well-known existing technique and will not be elaborated further. As a specific implementation, based on multiple sets of clear and blurry video frame samples at 1080P resolution, the distribution range of the Laplacian response variance is calculated for each set of samples. The intersection boundary of the minimum value of the clear sample and the maximum value of the blurry sample is then taken to determine the final sharpness threshold. For example, if the distribution range of the Laplacian response variance for clear samples is 1000-2500, and the distribution range for blurry samples is 100-1000, then the final sharpness threshold is determined to be 1000. Video frames with sharpness lower than the sharpness threshold are considered as uniformly bright video frames with unmet sharpness.

[0060] In other embodiments, the implementer may calibrate and adjust the sharpness threshold according to the video resolution.

[0061] This invention performs a sharpness and brightness uniformity analysis on all video frame images in the video frame sequence to be detected, and filters out all clear video frames that meet the criteria. This ensures that the video frame filtering results truly reflect the image quality level, effectively eliminates blurry frames and frames with abnormal exposure, avoids feature extraction failure due to image quality issues, and improves the accuracy of subsequent AI recognition.

[0062] Considering that personnel movement and equipment movement during construction can easily obscure key areas such as grouting holes and grout outlets, resulting in video data being unable to effectively reflect the grout status during obstruction and reducing the reliability of quality identification results, it is necessary to quantify the visibility integrity of each target hole area, determine the overall visibility of the video, and avoid AI identification misjudgments caused by obstruction.

[0063] Based on this, the specific implementation of the present invention includes: identifying the boundary contours of each target hole location region in each clear video frame, constructing a region visibility distribution matrix through grid division, calculating the local visibility integrity of each target hole location region, and determining the overall visibility of the acquired video.

[0064] In a specific embodiment of the present invention, the local visible integrity calculation process of each target hole location area is as follows: S31, identify the target hole location areas corresponding to the sleeve grouting hole, grout outlet hole and sealing hole in all clear video frames, and obtain the boundary position information of each target hole location area.

[0065] S32. Divide the boundary contour in the boundary location information of each target hole location area into several sub-region grids, count the occlusion coverage area of ​​each sub-region grid, and calculate the visibility coefficient of each sub-region grid by combining the boundary contour area of ​​the sub-region grid.

[0066] It should be noted that the visibility coefficient calculation process for each sub-region grid is as follows: obtain the difference between the boundary contour area of ​​the sub-region grid and the area of ​​the occlusion area of ​​the corresponding sub-region grid, and use the ratio of the area difference to the boundary contour area of ​​the sub-region grid as the visibility coefficient.

[0067] S33. Based on the spatial location of each sub-region grid, arrange the visibility coefficients of all sub-region grids to construct a regional visibility distribution matrix.

[0068] S34. Combine the preset spatial position weights of each sub-region grid to calculate the local visibility integrity of each target hole area.

[0069] Preferably, in a specific embodiment of the present invention, for example, the boundary contour of each target hole location region is divided into a 3×3 grid, and the weight of the central region of the hole location is higher than that of the edge region. For example, the weight of the central grid is 0.2, the weight of the edge grid is 0.1, and the sum of the weights is 1. The visibility coefficient of each sub-region grid is multiplied by the corresponding preset spatial position weight and then summed to obtain the local visibility integrity of each target hole location region.

[0070] In a specific embodiment of the present invention, the overall visibility determination process of the acquired video is as follows: if the local visibility integrity of a target hole area in a clear video frame is lower than the basic hole visibility integrity threshold (e.g., 0.7, to ensure that the central area is unobstructed), then the clear video frame is recorded as a hole-invisible video frame, and the target hole area is regarded as an invisible hole area.

[0071] If there are no consecutive invisible video frames for all holes (three or more video frames are defined as consecutive video frames), or if consecutive invisible video frames for holes correspond to different areas of invisible holes (non-same hole area continuously obscures the view), then the overall visibility of the captured video is deemed acceptable; otherwise, the overall visibility of the captured video is deemed unacceptable.

[0072] This invention identifies the boundary contours of each target borehole region in each clear video frame, constructs a regional visibility distribution matrix through grid division, calculates the local visibility integrity of each target borehole region, and determines the overall visibility of the acquired video. This enables a refined quantitative assessment of the visibility of target boreholes, effectively filters out videos with borehole visibility integrity that meet the requirements, ensures the reliability of the data foundation for judging the quality of grouting construction, and achieves effective identification and supervision of grouting construction procedures.

[0073] Considering that video stability directly determines the continuity of hole location feature extraction during AI recognition, and that existing technologies are prone to severe image shaking due to handheld shooting or interference from the construction environment, causing disorder in hole location feature extraction during AI recognition, it is necessary to quantify the translation and rotation of the video, construct video stability indicators, and achieve accurate assessment of the stability of grouting construction videos.

[0074] Based on this, the specific implementation of the present invention includes: acquiring the pixel motion vector field between all adjacent frames, extracting the overall translational motion and rotational motion between each adjacent frame, and forming a stability index for the acquired video.

[0075] In a specific embodiment of the present invention, the method for extracting the overall translational motion is as follows: S411, grayscale processing is performed on each clear video frame to obtain the position of all pixels, and the motion vector of each pixel between all adjacent frames is obtained by position comparison, thereby generating a pixel motion vector field between all adjacent frames.

[0076] S412. Extract the motion vector of each pixel from the pixel motion vector field and decompose it to generate the horizontal displacement component and the vertical displacement component of each pixel.

[0077] S413. Statistically analyze the horizontal and vertical displacement components of all pixels to determine the central tendency values ​​of the horizontal and vertical displacement components. The central tendency value is the median value, which is denoted as the overall horizontal translation and the overall vertical translation between frames, respectively.

[0078] S414. The overall horizontal translation amount between frames and the overall vertical translation amount between frames are vectorized to obtain the overall translational motion amount between adjacent frames.

[0079] In a specific embodiment of the present invention, the method for extracting the rotational motion is as follows: S421, establish a polar coordinate system with the center of the video frame as the origin, and divide it into several concentric annular regions according to different radius distances.

[0080] S422. Based on the motion vector of each pixel in the pixel motion vector field, obtain the radial and tangential components of the motion vector of each pixel, and calculate the average value of the tangential components of all pixels in each concentric annular region to obtain the average tangential velocity of each concentric annular region.

[0081] S423. Analyze the ratio of the average tangential velocity of each concentric annular region to the corresponding radius distance to obtain the local rotational angular velocity of each concentric annular region, and compare it to generate a video frame to estimate the rotational angular velocity. The estimated rotational angular velocity of the video frame is the median of the local rotational angular velocities of each concentric annular region.

[0082] S424. Multiply the estimated rotation angular velocity of the video frame with the time interval between adjacent frames to obtain the change in rotation angle between frames, and use it as the rotational motion between adjacent frames.

[0083] S425. The overall translational and rotational motion between adjacent frames is used as an indicator of the stability of the acquired video.

[0084] This invention acquires the pixel motion vector field between all adjacent frames, extracts the overall translational and rotational motion between adjacent frames, and constructs a video stability index to achieve accurate evaluation of the stability of grouting construction videos. It effectively eliminates videos with image jitter exceeding the recognition tolerance range, avoids the problem of disordered hole position feature extraction during AI recognition, and improves the accuracy of AI recognition for grouting construction.

[0085] Considering that video quality needs to be judged comprehensively based on visibility and stability, passing a single dimension cannot guarantee the AI ​​recognition effect. It is necessary to evaluate the overall quality based on the overall visibility and stability indicators, output the video admission result, and ensure that only qualified videos enter the subsequent AI recognition process.

[0086] Based on this, the specific implementation of the present invention includes: evaluating the overall quality of the acquired video based on the overall visibility and stability indicators of the acquired video, and outputting video admission results based on the overall quality of the acquired video.

[0087] In one specific embodiment of the present invention, the video admission result output content is as follows: when the overall visibility of the acquired video is unqualified, the overall quality of the acquired video is judged to be unqualified, and the video admission result is output as unqualified.

[0088] If the overall visibility of the captured video is acceptable, the stability index of the captured video is compared with the maximum allowable position threshold and the maximum allowable angle threshold corresponding to the video stability.

[0089] If any motion quantity in the stability index of the acquired video exceeds the corresponding threshold, the overall quality of the acquired video is judged to be unqualified, and the output video admission result is rejected.

[0090] Conversely, if the overall quality of the acquired video is deemed acceptable, the output video admission result will be deemed passed.

[0091] It should be noted that the implementation steps for setting the maximum allowable position threshold and the maximum allowable angle threshold for preset video stability are as follows: First, collect qualified grouting video samples and unqualified shaking video samples under different shooting conditions at the construction site. Calculate the pixel motion vector field for all samples, extract the overall translational motion and rotational motion between adjacent frames of each sample, and count the maximum value of translational motion and the maximum value of rotational motion in qualified samples. Use these as the initial threshold values.

[0092] Then, based on the overall translational and rotational motion of the unqualified jittery video samples, they are compared with the corresponding initial threshold values. The proportion of unqualified jittery video samples whose overall translational and rotational motion are both greater than the corresponding initial threshold values ​​is counted. If the proportion of unqualified jittery video samples does not reach 95%, the maximum allowable position threshold is fine-tuned with a step size of 0.5 pixels / frame and the maximum allowable angle threshold is fine-tuned with a step size of 0.2° / frame. The actual test is repeated until the accuracy of the judgment meets the requirements. Finally, the maximum allowable position threshold and the maximum allowable angle threshold are output.

[0093] This invention evaluates the overall quality of acquired videos based on the overall visibility and stability indicators of the acquired videos, and outputs video admission results to ensure that the acquired videos meet the technical requirements of AI intelligent recognition, improve the usability of the acquired videos, and provide reliable data support for the automated detection and traceability of grouting construction quality.

[0094] like Figure 3 As shown in another example, the present invention provides a video quality intelligent verification system for grouting construction, including a shooting angle acquisition module, a clear video frame screening module, a visibility determination module, a stability index composition module, and a video access assessment module.

[0095] The modules are connected as follows: the shooting angle acquisition module is connected to the clear video frame screening module; the visibility determination module is connected to both the clear video frame screening module and the stability index composition module; and the video admission assessment module is connected to the stability index composition module.

[0096] The shooting angle acquisition module extracts the video frame sequence to be detected from the video collected at the grouting construction site. Based on the component edge feature information in the sequence, it calculates the shooting angle between the component plane normal vector and the camera optical axis direction, and determines whether the shooting angle meets the grouting recognition requirements.

[0097] The clear video frame filtering module, when meeting the grouting identification requirements, performs a clarity and brightness uniformity analysis on all video frame images in the video frame sequence to be detected, and filters out all clear video frames that meet the requirements.

[0098] The visibility determination module identifies the boundary contours of each target hole location region in each clear video frame, constructs a regional visibility distribution matrix through grid division, calculates the local visibility integrity of each target hole location region, and determines the overall visibility of the acquired video.

[0099] The stability index module acquires the pixel motion vector field between all adjacent frames, extracts the overall translational and rotational motion between adjacent frames, and constitutes the stability index of the acquired video.

[0100] The video access assessment module evaluates the overall quality of the acquired video based on the overall visibility and stability indicators of the acquired video, and outputs the video access result based on the overall quality of the acquired video.

[0101] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0102] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0103] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0104] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0105] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A video quality intelligent verification method for grouting construction, characterized in that, include: Extract the video frame sequence to be detected from the video collected at the grouting construction site. Based on the component edge feature information in the sequence, calculate the shooting angle between the component plane normal vector and the camera optical axis direction, and determine whether the shooting angle meets the grouting recognition requirements. When the grouting identification requirements are met, the sharpness and brightness uniformity of all video frame images in the video frame sequence to be detected are analyzed, and all clear video frames that meet the requirements are selected. Identify the boundary contours of each target hole location region in each clear video frame, construct a region visibility distribution matrix through grid division, calculate the local visibility integrity of each target hole location region, and determine the overall visibility of the acquired video. Obtain the pixel motion vector field between all adjacent frames, extract the overall translational and rotational motion between each adjacent frame, and form the stability index of the acquired video. Based on the overall visibility and stability indicators of the acquired video, the comprehensive quality of the acquired video is evaluated, and the video admission result is output based on the comprehensive quality of the acquired video.

2. The intelligent video quality verification method for grouting construction according to claim 1, characterized in that: The step for determining whether the shooting angle meets the requirements for grouting identification is as follows: The video frame sequence to be detected is extracted at preset time intervals to obtain all video frames to be analyzed. Edge pixel detection is performed on each video frame to be analyzed, and the set of component edge line segments in all video frames to be analyzed is extracted. Straight line filtering is performed on the set of component edge line segments to determine the horizontal edge direction vector and vertical edge direction vector of the component. The normal vector of the component plane is obtained by vector cross product operation. Combined with the preset camera optical axis direction vector, the shooting angle between the normal vector of the component plane and the camera optical axis direction is calculated. If the shooting angle meets the preset criteria for grouting AI recognition, the shooting angle is deemed to meet the grouting recognition requirements; otherwise, the shooting angle is deemed not to meet the grouting recognition requirements, and the result of unqualified video acquisition is output.

3. The intelligent video quality verification method for grouting construction according to claim 2, characterized in that: The horizontal edge direction vector and vertical edge direction vector of the component are determined as follows: Perform line recognition on the component edge line segment set, and extract the corresponding horizontal and vertical edge line segments of the component; Obtain the length and direction of the horizontal and vertical edge segments of the component, and filter out interference segments based on the dimensions of the grouting sleeve corresponding to the grouting construction. The filtered horizontal and vertical edge segments are subjected to directional consistency analysis, and the horizontal and vertical baselines of the component are fitted to obtain the horizontal edge direction vector corresponding to the horizontal edge baseline and the vertical edge direction vector corresponding to the vertical baseline.

4. The intelligent video quality verification method for grouting construction according to claim 1, characterized in that: The steps for filtering out all clear video frames that meet the criteria are as follows: The images of each video frame in the video frame sequence to be detected are converted to grayscale. The brightness distribution characteristics of each video frame image are statistically analyzed based on the grayscale histogram. The video frame image is then judged to have abnormal exposure based on the brightness distribution characteristics. If a video frame does not have abnormal exposure, then the video frame is recorded as a uniformly bright video frame. Convolution operation is performed on all uniform brightness video frames to obtain the Laplacian response value of the uniform brightness video frames, and the sharpness of each uniform brightness video frame is obtained by variance calculation. Remove video frames with uniform brightness that do not meet the clarity standard, and count all clear video frames that meet the standard.

5. The intelligent video quality verification method for grouting construction according to claim 1, characterized in that: The calculation process for the local visible integrity of each target aperture region is as follows: Identify the target hole locations corresponding to the sleeve grouting holes, grout outlet holes, and sealing holes in all clear video frames, and obtain the boundary location information of each target hole location area; The boundary contours in the boundary location information of each target hole location area are divided into several sub-region grids. The occlusion coverage area of ​​each sub-region grid is counted, and the visibility coefficient of each sub-region grid is calculated by combining the boundary contour area of ​​the sub-region grid. Based on the spatial location of each sub-region grid, the visibility coefficients of all sub-region grids are arranged to construct a regional visibility distribution matrix; By combining the preset spatial location weights corresponding to each sub-region grid, the local visibility integrity of each target hole location region is calculated.

6. The intelligent video quality verification method for grouting construction according to claim 5, characterized in that: The process for determining the overall visibility of the captured video is as follows: If the local visibility integrity of a target hole location region in a clear video frame is lower than the basic hole location visibility integrity threshold, then the clear video frame is recorded as a hole location invisible video frame, and the target hole location region is regarded as an invisible hole location region. If there are no consecutive invisible video frames for all holes, or if consecutive invisible video frames for holes correspond to different areas of invisible holes, then the overall visibility of the acquired video is deemed acceptable; otherwise, the overall visibility of the acquired video is deemed unacceptable.

7. The intelligent video quality verification method for grouting construction according to claim 1, characterized in that: The method for extracting the overall translational motion is as follows: Each clear video frame is processed into grayscale to obtain the position of all pixels. The motion vectors of each pixel between all adjacent frames are obtained by position comparison, and a pixel motion vector field between all adjacent frames is generated. The motion vectors of each pixel are extracted from the pixel motion vector field and decomposed to generate the horizontal and vertical displacement components of each pixel. The horizontal and vertical displacement components of all pixels are statistically analyzed to determine the central tendency values ​​of the horizontal and vertical displacement components, which are denoted as the overall horizontal translation and the overall vertical translation between frames, respectively. The overall horizontal translation between frames and the overall vertical translation between frames are vectorized to obtain the overall translational motion between adjacent frames.

8. The intelligent video quality verification method for grouting construction according to claim 7, characterized in that: The method for extracting rotational motion is as follows: A polar coordinate system is established with the center of the video frame as the origin, and the system is divided into several concentric annular regions according to different radius distances. Based on the motion vector of each pixel in the pixel motion vector field, obtain the radial and tangential components of the motion vector of each pixel, and calculate the average tangential velocity of each concentric ring region by averaging the tangential components of all pixels in each concentric ring region. By analyzing the ratio of the average tangential velocity of each concentric annular region to the corresponding radius distance, the local rotational angular velocity of each concentric annular region is obtained, and the rotational angular velocity is estimated by comparing the generated video frame. The estimated rotation angular velocity of the video frame is multiplied by the time interval between adjacent frames to obtain the change in rotation angle between frames, which is then used as the rotational motion between adjacent frames. The overall translational and rotational motion between adjacent frames is used as an indicator of the stability of the acquired video.

9. The intelligent video quality verification method for grouting construction according to claim 8, characterized in that: The video admission result output content is as follows: If the overall visibility of the captured video is unqualified, the overall quality of the captured video is judged to be unqualified, and the video admission result is rejected. If the overall visibility of the captured video is acceptable, the stability index of the captured video is compared with the maximum allowable position threshold and the maximum allowable angle threshold corresponding to the video stability. If any motion quantity in the stability index of the acquired video exceeds the corresponding threshold, the overall quality of the acquired video is judged to be unqualified, and the video admission result is rejected. Conversely, if the overall quality of the acquired video is deemed acceptable, the output video admission result will be deemed passed.

10. A video quality intelligent verification system for grouting construction, characterized in that: The shooting angle acquisition module extracts the video frame sequence to be detected from the video collected at the grouting construction site. Based on the component edge feature information in the sequence, it calculates the shooting angle between the component plane normal vector and the camera optical axis direction, and determines whether the shooting angle meets the grouting recognition requirements. The clear video frame filtering module performs a clarity and brightness uniformity analysis on all video frame images in the video frame sequence to be detected when the grouting recognition requirements are met, and filters out all clear video frames that meet the requirements. The visibility determination module identifies the boundary contours of each target hole location region in each clear video frame, constructs a region visibility distribution matrix through grid division, calculates the local visibility integrity of each target hole location region, and determines the overall visibility of the acquired video. The stability index module acquires the pixel motion vector field between all adjacent frames, extracts the overall translational and rotational motion between each adjacent frame, and constitutes the stability index of the acquired video. The video access assessment module evaluates the overall quality of the acquired video based on the overall visibility and stability indicators of the acquired video, and outputs the video access result based on the overall quality of the acquired video.

Images